Agent skill

Second Brain Fix

by coleam00 in coleam00/skills

Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left…

MITAuto-check passedKnowledge Management

Install Second Brain Fix

skills CLI
$ npx skills add coleam00/skills --skill second-brain-fix -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install coleam00/skills second-brain-fix --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/coleam00/skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/second-brain-fix .claude/skills/second-brain-fix && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
second-brain-fix
GitHub stars
670
Token cost
~3.4k tokens
SKILL.md length
2,016 words
Files
1
Skills in repo
34
Repo updated
First seen
Licence
MIT

At a glance

Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left…

  • Works in 6 steps: build the batch list → the contradicted pile, in batch → convert the locations, in batch → …
  • Tasks that involve Second brain
  • SKILL.md covers The argument, This skill writes in bulk.…, The three piles are three… and How much to take on, plus 8 more sections
  • Calls git

What it does

Second Brain Fix is an agent skill from coleam00/skills. Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left behind. Use after running second-brain-audit, when someone wants to fix the rest of their notes rather than one page, when they ask how to take the audit forward, when they have a list of contradicted or unsupported claims to work through, or when a notes folder needs converting in bulk rather than one page at a time.

Its SKILL.md is about 3.4k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Knowledge Management, covering Second brain. The repository describes itself as: The agent skills I actually use to build software with coding agents. The PIV loop, planning, worktrees, and the meta-skills for building your own AI Layer. The licence is MIT.

When your agent uses it

  • Tasks that involve Second brain

Example prompts

  • “/second-brain-fix”

Workflow steps

6 steps, taken from the step headings in SKILL.md.

  1. build the batch list
  2. the contradicted pile, in batch
  3. convert the locations, in batch
  4. the unsupported pile, one pass of questions
  5. write the results back
  6. re-audit and compare

What it can do on your machine

Read from SKILL.md and the folder at commit 847be08. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Second Brain Fix loads about 3.4k tokens when it runs. Until then it costs about 133 tokens; SKILL.md has 2,016 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~133
When it runs · the whole SKILL.md, loaded when a task matches
~3.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from coleam00/skills at commit 847be08, republished under its MIT licence (© coleam00). 2,016 words, ~3,387 tokens.

Download SKILL.mdSave it as .claude/skills/second-brain-fix/SKILL.md (or your agent's skills folder).
name
second-brain-fix
description
Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left behind. Use after running second-brain-audit, when someone wants to fix the rest of their notes rather than one page, when they ask how to take the audit forward, when they have a list of contradicted or unsupported claims to work through, or when a notes folder needs converting in bulk rather than one page at a time.
argument-hint
[path-to-audit-findings.md]
arguments
audit-results

Second Brain Fix

The audit found what is wrong and fixed one location so the shape was visible. This works through the rest, in batches, from the findings the audit left behind.

The argument

The findings file this run was invoked with is $audit-results, and it is optional:

/second-brain-fix                                   nothing passed
/second-brain-fix ~/notes                           a folder to look in
/second-brain-fix ~/notes/second-brain-audit.md     the findings file itself

$audit-results is a path to one markdown file of findings. /second-brain-audit names what it writes second-brain-audit.md, so that is usually what you will be handed, but the name is not the contract. Any markdown file holding the findings works, and whatever the user points at is the one to use.

Resolve it before anything else:

What you were givenDo
a path to a filethat is the findings. The notes folder is its parent, unless the file names a different one
a path to a folderlook inside it for the findings, second-brain-audit.md first
nothing, so the line above still reads \$audit-resultslook in the current folder, then one level down
a path that does not existsay so and stop. Do not fall back to searching, or a typo silently works on the wrong folder
more than one candidate when searchinglist them with their newest ## Log dates and ask which
no candidate when searchingstop. Say to run /second-brain-audit first, since there is nothing to work from

Never invent a queue from scratch when there are no findings. Auditing and fixing in one pass is how a bulk write happens against findings nobody read.

The findings are the input. $audit-results is both the queue and the record: one keyed line per location in ## Current State, one dated line per run in ## Log. Everything below is driven by that file, and a run that does not update it is a run nobody can pick up from.

This skill writes in bulk. Settle backups before it does.

Never start writing without settling this, and never resolve it by copying the notes folder somewhere else. A duplicate folder of notes is a second answer to every question, which is the exact problem this skill exists to remove.

First, check what is already there. Run git rev-parse --show-toplevel in the notes folder.

  • Already in git, either its own repo or tracked by a parent one. Commit or stash anything outstanding, then git checkout -b second-brain-fix, and commit after every batch. Say which repo you are in, because a vault nested inside another project is common and the user should know what the branch covers.

  • Not in git. Ask, in these words or close to them:

    Your notes are not in git, so there is no way to undo a batch. Want me to run git init here first? It stays on your machine, nothing gets pushed anywhere, and every batch becomes a checkpoint you can roll back to.

    If they agree: git init, git add -A, one baseline commit, then the branch. Never add a remote and never push.

    If they decline, ask once more and take the answer:

    Understood. To be clear, this rewrites many files at once and there will be no way to undo it. Go ahead anyway?

    On a clear yes, proceed and do not raise it again. On silence or anything ambiguous, stop. An unanswered question is not permission.

The three piles are three different jobs

PileWhat the finding containsWhat to do
Contradictedthe stale claim AND the newer evidencefix in batch, no questions
Unsupportedthe claim, and nothing backing it upnever guess. One batch of questions for the user
Locations not yet converteda page that has no ## Current State / ## Logconvert in batch, verbatim

The middle row is where bulk fixing goes wrong. An unsupported claim has no answer in the notes by definition, so an agent told to "correct these" will invent a value or quietly delete the line. Both are worse than the stale claim.

How much to take on

Ask how large the notes are, or count the markdown files, and scale the ambition:

SizeDo this
Under ~50 fileseverything in one session. Two or three batches
~50 to 500the always-loaded surface first, then one folder per batch, committing between
500+the always-loaded surface, plus only the pages with a trail (several entries about one subject over time). Everything else stays as it is, permanently

Bigger notes do not mean convert more. They mean convert a smaller fraction and lean harder on the write path, because at that size nobody is ever going to hand-tend the archive.

Step 1: build the batch list

Read $audit-results. Take the ## Current State entries that are not marked as fixed, plus any pile the audit reported but did not enumerate. Group them into batches of roughly 5 to 15 locations, keeping a folder together where you can.

Show the user the batch list and the order before writing anything. Ordering is always:

  1. Whatever loads every session.
  2. The pages the agent gets wrong most often.
  3. Pages with a trail.
  4. Nothing else.

Step 2: the contradicted pile, in batch

For each finding, the newer evidence is already named. So:

  • Replace the stale line in ## Current State.
  • Move the superseded line, verbatim, to ## Log.
  • Date the new line with the date of the evidence, not today.

That last one matters more than it looks. A date on a current value is a claim that somebody checked the value on that day. Stamping today's date on a line you corrected from a note written in June makes a June fact look verified this morning.

Work the whole batch, then show one summary: how many lines replaced, in which files, and the three or four that were least obvious. Do not show a diff per line, and do not ask per line.

Step 3: convert the locations, in batch

Give each page the two sections. Adapt to the shape the audit found:

markdown
## Current State
<!-- One entry per subject. Dated. REPLACED on update, never appended to. -->

- **Retainer** (2026-08-01): $3,200/mo, renewed through February 2027
- **Main contact** (2026-05-02): Curtis Ilo

## Log
<!-- Append-only. Never edit or delete an entry. -->

- (2026-04-30) Delivered and paid, $21,000
- (2026-06-15) Added reply drafting, retainer to $3,200/mo

One big file gets a ## Current State block at the top and everything else beneath it, no new files. Daily notes get one new file of current values and the journal untouched. Notes that are not markdown do not get converted at all.

This is sorting, not rewriting. Every existing line lands in one of the two sections, verbatim, at most with a date prepended. Improving the prose is how information disappears without anyone noticing, and in a batch nobody is reading closely enough to catch it.

When a bullet has no date and none can be recovered, do not invent one and do not quietly leave the entry bare. Take the date from the evidence if the bullet cites any, otherwise write (date unknown) in the date's place. A bare entry reads as an oversight; (date unknown) reads as the gap it is, and it survives into the next audit as something to ask about.

Two kinds of page to leave alone, and say so rather than converting them:

  • Reference checklists. Packing lists, hospital-bag lists, standard operating steps. Few bolded keys, few dates, and the order is the content. Converting one passes every structural check and destroys the thing that made it useful.
  • Pages where the freshest status lives inside prose, under a heading that owns the bullets below it. Dissolving that section leaves the newest status outside ## Current State and the oldest inside it, which is the exact failure being removed, rebuilt one level up.
Show full SKILL.md (789 more words)Show less

Step 4: the unsupported pile, one pass of questions

Collect them all and ask once, as a numbered list. Not one at a time, and never silently.

For each, three outcomes:

  • Still true goes into ## Current State with the date the user gives, and say plainly that the evidence was missing, not just misfiled.
  • No longer true gets replaced, old line verbatim to ## Log.
  • Cannot tell comes out of the always-loaded file entirely. A confident wrong answer costs more than a missing one.

If the user does not answer, leave every one of them exactly as it is. An unanswered question is not permission.

Anything you leave for the owner gets annotated where it lives

A question in your summary reaches one person once. The line stays in the notes and gets read by every future session, so the warning has to be on the line, not only in your report.

Whenever you leave a line standing because only the owner can resolve it, append the reason to that line in place, keeping the original text intact ahead of it:

markdown
- **Sam Iwu**: accepted a full-time offer, starting October 2026. UNVERIFIED (fix run
  2026-09-18): no date and no source anywhere; entities/sam-iwu.md says he is still a
  contractor. Confirm or delete.

- **Ferro retainer renewal** (2026-09-12): renewed at $6,500/mo through year-end. CONFLICT
  (fix run 2026-09-18): the 2026-09-10 call set it to $12,000/mo. Nothing sources this line.
  Ask before quoting either number.

This matters most for a stale line that is newer than the line you just corrected. If one key says $12,000 dated the 10th and another says $6,500 dated the 12th, an agent reading that page takes the newer one and answers $6,500, so correcting the first key changed nothing unless the second one carries the warning. Check for that case explicitly after every batch: for each key you fixed, look for another key on the same page about the same subject with a later date.

Never delete the line to resolve the conflict, and never pick a value yourself. Annotating is the whole move.

Step 5: write the results back

After each batch, edit $audit-results:

  • Replace each location's ## Current State line with its new status. One line per location, always. Never a second line.
  • Record every line you annotated and left standing, so the next run can see what is still waiting on the owner rather than re-deriving it.
  • Append one entry to ## Log: the date, which batch, how many lines replaced, how many locations converted, and how many unsupported claims are still unanswered.

Then commit the batch. The findings and the notes move together, so an interrupted run is resumable by reading one file.

Step 6: re-audit and compare

When the batches are done, run /second-brain-audit again and compare the contradicted count to the last ## Log line. That number is the only evidence any of this worked.

If the count did not move much, say so and say why rather than presenting the conversion as a result. Restructuring cannot reach a fact nobody ever wrote down, and when the count stays flat that is usually what happened. The fix then is the write path, not another batch.

Flag one thing if the findings file has been renamed: audit.py skips files whose name starts with second-brain-audit, so a renamed one gets read as evidence on the next scan, and it quotes stale claims verbatim. Either keep that name or keep the file outside the notes folder.

Rules that never bend

  1. Lose nothing. Every line lands somewhere, verbatim. If a line cannot be placed, leave it where it is and report it.
  2. Never merge two subjects that look alike. "Acme (May)" and "Acme Corp renewal" may be two real things. A duplicate entry is a cheap mistake; a wrong merge destroys information. Report near-misses at the end of the batch and let the user decide. Batch work merges by default because merging looks like tidying, so this rule needs holding on purpose.
  3. Never invent a value, and never delete a claim to make a pile smaller. Anything you leave for the owner carries its reason on the line itself, not only in your summary. A warning that lives in a chat window is the failure this skill exists to fix.
  4. Never edit or delete a ## Log entry. Old and superseded is the point of that section.
  5. A date is a claim that the value was checked. A line you only moved keeps its own date.
  6. Never copy the notes folder as a backup. Git, or an informed no. Nothing else.
  7. Stop when the always-loaded surface has nothing contradicted. An archive full of old pages is history, not rot. There is no version of this where every page gets converted.

If the write path has not changed yet

Check whether CLAUDE.md (or AGENTS.md, or the system prompt) carries the state/event rule from the audit's phase 7. If it does not, add it before starting, and say why: fixing four hundred lines under a write path that can only append buys about a month.

© coleam00, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/second-brain-fix of coleam00/skills.

Open the folder on GitHubat commit 847be08

Compare with similar skills

Second Brain Fix next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Second Brain Fix compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Second Brain Fix this skillcoleam00/skills670—~3.4kAutomated safety check: PassMIT
Second BrainNicholasSpisak/second-brain7361 repos~1.5kAutomated safety check: NotesNone
3D Brainnateherkai/AIS-OS1.6k—~2.1kAutomated safety check: PassCustom licence
Obsidian Canvas BoardsAgriciDaniel/claude-obsidian15k—~1.4kAutomated safety check: PassMIT
Icm ArchitectRinDig/icm-architect1.8k—~3.2kAutomated safety check: PassMIT
Obsidian CLI Read TransportAgriciDaniel/claude-obsidian15k1 repos~762Automated safety check: PassMIT

Similar skills

  • Second Brain

    NicholasSpisak/second-brain

    Set up a new Obsidian knowledge base with the LLM Wiki pattern.

    736 GitHub starsUsed in 1 repo~1.5k tokens
    Knowledge ManagementAuto-check: notes
  • 3D Brain

    nateherkai/AIS-OS

    A skill your agent uses when someone asks to build a 3D brain, visualize their AIOS or second brain, turn their knowledge into an interactive graph, or run /3d-brain or /3D brain.

    1.6k GitHub stars~2.1k tokensUpdated 1 mo ago
    Knowledge ManagementAuto-check passed
  • Obsidian Canvas Boards

    AgriciDaniel/claude-obsidian

    Creates, inspects and updates Obsidian JSON Canvas boards in a vault, with text, file, link, group and edge nodes, using safe recoverable edits.

    15k GitHub stars~1.4k tokensUpdated 28 days ago
    Knowledge ManagementAuto-check passed
  • Icm Architect

    RinDig/icm-architect

    Design any process, idea, problem, or body of knowledge into an ICM (Interpretable Context Methodology) workspace — folder structure as agent architecture — or restructure an existing folder, repo…

    1.8k GitHub stars~3.2k tokensUpdated 1 mo ago
    Knowledge ManagementAuto-check passed
  • Obsidian CLI Read Transport

    AgriciDaniel/claude-obsidian

    Detects and uses the official Obsidian command-line interface for read-only vault access, falling back to direct file reads, while all mutations stay on a separate transaction core.

    15k GitHub starsUsed in 1 repo~762 tokens
    Knowledge ManagementAuto-check passed
  • Cascading Goal Tracker

    ballred/obsidian-claude-pkm

    Tracks three-year, yearly, monthly, and weekly goals in linked Markdown files, calculating completion percentages and flagging stalled goals.

    1.9k GitHub stars~1.5k tokensUpdated 7 mo ago
    Knowledge ManagementAuto-check passed

More from coleam00/skills

All 34 skills in this repo
  • Ablate AI Layer

    coleam00/skills

    Measure whether a repository's AI instructions still earn their place, by running the same real task many times with the layer intact and with it stripped, then grading every rule against what…

    670 GitHub stars~1.9k tokensUpdated yesterday
    Auto-check passed
  • Build Dark Factory

    coleam00/skills

    Take a PRD and build a dark factory around it - a repository that takes work in as an issue and ships validated code out with nobody at the keyboard - one component at a time, into the user's actual…

    670 GitHub stars~12k tokensUpdated yesterday
    Auto-check passed
  • Drive Screen

    coleam00/skills

    Take real control of the desktop - list and focus windows, type, paste, click, scroll, and screenshot - on Windows, macOS or Linux, and drive other coding-agent sessions running in terminals.

    670 GitHub stars~5.4k tokensUpdated yesterday
    Auto-check passed
  • Second Brain Audit

    coleam00/skills

    Audit any second brain, notes folder, or agent memory for facts that have quietly stopped being true, then fix the worst one so it stops recurring.

    670 GitHub stars~4.6k tokensUpdated yesterday
    Auto-check passed
  • Build Signal Engine

    coleam00/skills

    Build a personal signal engine from scratch - a system that reads every source someone cares about each day (changelogs and release notes, communities, feeds, videos, papers), makes a quick decision…

    670 GitHub stars~2.9k tokensUpdated yesterday
    Auto-check passed
  • Worktree Create

    coleam00/skills

    Create one or more git worktrees for parallel development, each on its own branch with gitignored config copied in, dependencies installed, and a health check, by fanning out a setup subagent per…

    670 GitHub stars~958 tokensUpdated yesterday
    Auto-check passed

Questions about Second Brain Fix

What does Second Brain Fix do?

Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left…. Second Brain Fix is an agent skill from coleam00/skills. Work through the findings from a second-brain audit in batches, correcting stale facts and converting locations onto the state/event schema, driven by the markdown file of findings the audit left behind.

When should I use Second Brain Fix?

Second Brain Fix fits situations like: tasks that involve Second brain.

How do I install Second Brain Fix in Claude Code?

Run `npx skills add coleam00/skills --skill second-brain-fix -a claude-code`. Or copy the skill folder (.claude/skills/second-brain-fix in coleam00/skills) into .claude/skills/second-brain-fix in your project. Claude Code loads it when a task matches its description.

How do I install Second Brain Fix in Codex?

Run `npx skills add coleam00/skills --skill second-brain-fix -a codex`. Or copy the skill folder (.claude/skills/second-brain-fix in coleam00/skills) into .agents/skills/second-brain-fix in your project. Codex loads it when a task matches its description.

Can I use Second Brain Fix in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add coleam00/skills --skill second-brain-fix -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/second-brain-fix, .gemini/skills/second-brain-fix, .github/skills/second-brain-fix and .opencode/skills/second-brain-fix in your project.

What does Second Brain Fix need to run?

Going by SKILL.md and its folder, Second Brain Fix needs the command-line tools its instructions call (git).

Does Second Brain Fix access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Second Brain Fix safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Second Brain Fix use?

Second Brain Fix is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Second Brain Fix use?

About 3.4k tokens (SKILL.md is roughly 14k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Second Brain Fix?

Skills that share tags, products or a category with Second Brain Fix: Second Brain (NicholasSpisak/second-brain, 736 stars), 3D Brain (nateherkai/AIS-OS, 1.6k stars), Obsidian Canvas Boards (AgriciDaniel/claude-obsidian, 15k stars) and Icm Architect (RinDig/icm-architect, 1.8k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Second Brain Fix?

coleam00 (a GitHub user) maintains it in coleam00/skills, which has 670 GitHub stars. The repository holds 34 skills in this directory. The repository was last updated on October 7, 2026.

Source: coleam00/skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.